openai.com: This technical report focuses on (1) our method for turning visual data of all types into a unified representation that enables large-scale training of generative models, and (2) qualitative evaluation of Sora’s capabilities and limitations. Model and implementation details are not included in this report.
3 comments:
This was a very technical read. I found that also illuminating, because it came directly from OpenAI, as opposed to a third party covering the things they were talking about. A lot of sources try to essentially rewrite the story in a way that doesn’t cover everything, and it’s nice seeing a primary source used for something like this. A big thing that this focus on was the very broad strokes about how this kind of generation and data analysis is done, something which I didn’t really have a firm idea of. It’s still very technical, but it’s nice seeing a bit into the inner workings of a system, especially with what the new AI technology could mean for design. What was interesting was seeing the broad spectrum of ways that this new technology can actually create, not just simple prompt based things, but everything else as well, from extension to cleanup. That might become incredibly useful in the future, though it also could be very bad for the industry.
I found this report fascinating and impressive because of the level of detail that it went into. Many articles written about technology are meant to be for the average consumer and they are also written by an outside news outlet, so they will lack lots of the smaller details. OpenAI creating this report essentially ensures its accuracy and lets the consumer know all of the things that its technology is capable of. I think that this is particularly vital for a company like OpenAI which is on the forefront of innovation. They create the technology, and many people may not know all the uses for it without the developers telling them potential ideas as they do here. I was also glad that OpenAI admitted the parts of their generative video technology that are not yet up to par, instead of attempting to hide them and pretend that they do not exist.
I was really excited about this article because I thought it was about procedural generation in video games and how it works which is something I’m learning how to do right now. I probably should’ve realized that the article was from OpenAI and wouldn’t be about that but here we are. I have beef with openAI. I have followed their work since 2019. My best friend was a beta tester for GPT3. Up until ChatGPT became profitable, all their work was open source, hence the name OpenAI. It was in their mission and their motto. They believed free access to technology was the best way to technological advancement. They used many open source software in their own models. Then suddenly it became profitable and they kept their name but changed their mission. The generation of environments is cool enough. I think it’s a great tool for artists to use. If you generate a base environment as a reference and then work from there I think it’s great. However, I think this will end up being used to take away jobs from artists, which sucks.
Post a Comment